Mailing-List: contact dev-help@couchdb.apache.org; run by ezmlm
Precedence: bulk
Reply-To: dev@couchdb.apache.org
Received-SPF: neutral (nike.apache.org: local policy)
Content-Type: text/plain; charset=us-ascii
Mime-Version: 1.0 (Apple Message framework v1078)
Subject: Re: _changes timeout has a max of 60 seconds?
From: Damien Katz <damien@apache.org>
In-Reply-To: <4BE3A40E.9010502@skippinet.com.au>
Date: Fri, 7 May 2010 09:51:54 -0700
Content-Transfer-Encoding: quoted-printable
Message-Id: <5A8D3D77-D13C-499E-8CF6-402DE6C3D137@apache.org>
References: <4BE3A40E.9010502@skippinet.com.au>
To: dev@couchdb.apache.org

This is the intent.

The max timeout value is set in the ini. The reason we do this is to =
prevent lots of undetected dead connections using up resources, which =
can happen if the TCP connection dies abnormally.

To get around the timeout, you up the changes_timeout in the ini:

[httpd]
changes_timeout=3D#millisecs

Or you have the server send a heartbeat. Then the connection can stay =
open indefinitely, and the heartbeats cause dead connections to be =
detected by the server.

-Damien


On May 6, 2010, at 10:24 PM, Mark Hammond wrote:

> I'm trying to use _changes with feed=3Dcontinuous and a timeout =
greater than the default - however, couch always uses the default of 60 =
seconds.  Timeouts smaller than the default work as expected.  It seems =
get_changes_timeout/2 in couch_changes.erl is responsible for this by =
selecting the smallest of the specified and default timeout.
>=20
> While is sounds reasonable to have a default of 60 seconds, there =
doesn't seem to be an obvious case for also making that the max - *some* =
maximum might be reasonable, but 60 seconds seems pretty low.  Should I =
open a bug, or is this really the intent?  If the latter, I'll update =
the wiki accordingly...
>=20
> Thanks,
>=20
> Mark